Articles

E ffi cacy, safety, and immunology of an inactivated alum-adjuvant enterovirus 71 vaccine in children in China: a multicentre, randomised, double-blind, placebo-controlled, phase 3 trial

Feng-Cai Zhu, Fan-Yue Meng, Jing-Xin Li, Xiu-Ling Li, Qun-Ying Mao, Hong Tao, Yun-Tao Zhang, Xin Yao, Kai Chu, Qing-Hua Chen, Yue-Mei Hu, Xing Wu, Pei Liu, Lin-Yang Zhu, Fan Gao, Hui Jin, Yi-Juan Chen, Yu-Ying Don g, Yong-Chun Liang, Nian-Min Shi, Heng-Ming Ge, Lin Liu, Sheng-Gen Chen, Xing Ai, Zhen-Yu Zhang, Yu-Guo Ji, Feng-Ji Luo, Xiao-Qin Chen, Ya Zhang, Li-Wen Zhu, Zheng-Lun Liang, Xin-Liang Shen

Summary Lancet 2013; 381: 2024–32 Background A vaccine for enterovirus 71 (EV71) is needed to address the high burden of disease associated with Published Online infection. We assessed the effi cacy, safety, immunogenicity, antibody persistence, and immunological correlates of an May 29, 2013 inactivated alum-adjuvant EV71 vaccine. http://dx.doi.org/10.1016/ S0140-6736(13)61049-1 Methods We did a randomised, double-blind, placebo-controlled, phase 3 trial. Healthy children aged 6–35 months See Comment page 1968 from four centres in China were randomly assigned (1:1) to receive vaccine or alum-adjuvant placebo at day 0 and 28, Jiangsu Provincial Center for Disease Control and Prevention, according to a randomisation list (block size 30) generated by an independent statistician. Investigators and Nanjing, Jiangsu, China participants and their guardians were masked to the assignment. Primary endpoints were EV71-associated hand, (F-C Zhu MSc, F-Y Meng MSc, foot, and mouth disease (HFMD) and EV71-associated disease during the surveillance period from day 56 to month 14, J-X Li MSc, H Tao BS, K Chu MSc, analysed in the per-protocol population. This study is registered with ClinicalTrials.gov, number NCT01508247. Y-M Hu BS); Department of Clinical Microbiology and Immunology, College of Findings 10 245 participants were enrolled and assigned: 5120 to vaccine versus 5125 to placebo. 4907 (with three cases Pharmacy, Third Military of EV71-associated HFMD and eight cases of EV71-associated disease) versus 4939 (with 30 cases of EV71-associated Medical University & National HFMD and 41 cases of EV71-associated disease) were included in the primary effi cacy analysis. Vaccine effi cacy was Engineering Research Center for Immunological Products, 90·0% (95% CI 67·1–96·9) against EV71-associated HFMD (p=0·0001) and 80·4% (95% CI 58·2–90·8) against Chongqing, China (F-C Zhu, EV71-associated disease (p<0·0001). Serious adverse events were reported by 62 of 5117 (1·2%) participants in the J-X Li); Beijing Vigoo Biological, vaccine group versus 75 of 5123 (1·5%) in the placebo group (p=0·27). Adverse events occurred in 3644 (71·2%) Beijing, China (X-L Li MSc, versus 3603 (70·3%; p=0·33). Y-T Zhang PhD, Q-H Chen MSc, X-L Shen MSc); National Institute for Food and Drug Interpretation EV71 vaccine provides high effi cacy, satisfactory safety, and sustained immunogenicity. Control, Beijing, China (Q-Y Mao PhD, X Yao PhD, Funding China’s 12–5 National Major Infectious Disease Program, Beijing Vigoo Biological. X Wu MSc, F Gao MSc, Z-L Liang PhD); School of Public Health, Southeast University, Introduction the effi cacy and safety of the vaccine for prevention of Nanjing, Jiangsu, China Disease associated with enterovirus 71 (EV71) infection disease associated with EV71. (P Liu PhD, H Jin PhD, was fi rst described by Schmidt and colleagues in 1974,1 Y-J Chen MSc, Y-Y Dong MSc); Lianyungang City Center for who reported 20 patients with CNS disease—including Methods Disease Control and Prevention, one who died—in California, USA, between 1969, Study design and participants Lianyungang, Jiangsu, China and 1972. Since then, outbreaks of hand, foot, and We did this randomised, double-blind, placebo-controlled (L-Y Zhu BS); Baoying County mouth disease (HFMD) associated with EV71 infec- phase 3 trial in children aged 6–35 months at four centres Center for Disease Control and 2–5 Prevention, Yangzhou, Jiangsu, tion have been reported world wide —particularly in in China ( Donghai, Pizhou, and Baoying counties in China (Y-C Liang BS, S-G Chen BS, infants and young children—accounting for more than Jiangsu province, and Chaoyang district in Beijing) Y-G Ji BS, L-W Zhu BS); Chaoyang 6 000 000 cases and more than 2000 deaths in the past covering 32 townships and 485 villages from January, District Center for Disease decade.6 EV71 infection is associated with various 2012, to March, 2013. Eligibility was assessed by medical Control and Prevention, Beijing, China (N-M Shi MSc, X Ai MSc, diseases including HFMD, herpangina, neurological history inquiry and physical examination. Exclusion F-J Luo MSc); Donghai County signs (aseptic meningitis or encephalitis) with or without criteria included: history of HFMD, previous vaccination Center for Disease Control and serious sequelae, and nonspecifi c illnesses—eg, febrile with EV71 vaccine, acute febrile disease at recruitment, Prevention, Lianyungang, illness, viral exanthem, respiratory infection.6–8 A vaccine and taking any immunosuppressive drugs, blood Jiangsu, China (H-M Ge BS, Z-Y Zhang BS, X-Q Chen BS); and for EV71 is urgently needed to prevent and control products, drugs under investigation, inactivated vaccines, Pizhou County Center for epidemics of EV71.9,10 or attenuated live vaccines within a defi ned period Disease Control and Prevention, Phase 1 and phase 2 trials have assessed the safety (appendix p 3 shows full criteria). The proportion of Xuzhou, Jiangsu, China (L Liu BS, Y Zhang BS) and immunogenicity of an inactivated alum-adjuvant enrolled children in every township was less than 40% of EV71 vaccine in adults, children, and infants.5,11,12 We did the total number of children aged 6–35 months, to avoid this phase 3 trial in young children and infants to assess creating herd immunity in study areas (appendix p 4).

2024 www.thelancet.com Vol 381 June 8, 2013 Articles

The study was approved by the institutional review epidemiological, and laboratory data to confi rm EV71- Correspondence to: board of Jiangsu Provincial Center of Disease Control associated disease, and serious adverse events attributed Feng-Cai Zhu, Jiangsu Provincial and Prevention, and done in accordance with the to vaccine. The committee was not involved in any other Center for Disease Control and Prevention, Jiangsu Road, Declaration of Helsinki, Good Clinical Practice, and part of the study. Nanjing 210009, Jiangsu, China Chinese regulatory requirements. All guardians of Participants were visited at home or came to the clinic [email protected] participants provided written informed consent. The at 5, 8, 11, and 14 months, when any medically signifi cant or study protocol is available online. events were recorded. Digital photographs were used to Zheng-Lun Liang, National identify participants throughout immunisation, blood Institute for Food and Drug Randomisation and masking collection, and site visits. Control, Tiantanxili, Beijing 100050, China Three consecutive batches of EV71 vaccine (numbers We did active surveillance of EV17-associated disease in [email protected] 201108002, 201108003, and 201108004) and one batch of all 485 clinics in villages, all 32 hospitals in townships, and or placebo (number 201108001) were used for this study in a the seven biggest public hospitals in cities, of all vaccinated Xin-Liang Shen, Beijing Vigoo 1:1:1:3 ratio. Each dose of vaccine and placebo was assigned participants from day 56 to month 14. Guardians were Biological, Sanjianfangnanli, a code from a randomisation list by block randomisation encouraged to take their children to treatment in clinics or ChaoYang District, Beijing (block size 30). The randomisation list was generated by hospitals for any illness during the surveillance period. 100024, China an external statistician with SAS (version 9.1). All pack- Unwell participants were screened for any of: HFMD-like [email protected] ages of vaccine and placebo were identical in appearance, illness, herpangina, acute respiratory symptoms or gastro- See Online for appendix with the code as the only identifi er. Each participant was intestinal symptoms, with or without other systemic For the protocol see http:// assigned a sequential number according to their sequence compli cations—eg, CNS complications and cardio pul- www.jscdc.cn/jkzt/jbkzzt/ of enrolment and received vaccine or placebo labelled with monary failure—but excluding those who had disease that szkb/jsgf_yg/201305/ the same numbers. Thus, participants were randomly was confi rmed not to involve EV71. All potential cases were t20130517_35229.html assigned to receive EV71 vaccine or placebo in a 1:1 ratio. reported by the clinics or hospitals. Throat or rectal swabs Investigators involved in randomisation and masking did or both were taken for pathogen detection within 24 h. The not participate in any other part of the trial. Allocation ad hoc local laboratories in each study centre—operated by was masked from all participants, their guardians, Jiangsu Mole Bioscience— tested for Coxsackie A virus 16, and investigators. EV71, and other enteroviruses by real-time PCR (fl uores- cence assay) with a viral RNA diagnostic kit. Participants Procedures who were positive for EV71-specifi c RNA were deemed Inactivated alum-adjuvanted EV71 vaccine (vero cell), suspected cases and were visited by study staff s to assess containing 320 U of antigen and 0·18 mg of alum, was exposure history and take acute blood samples, serial developed and manufactured by Beijing Vigoo Biological swabs, and stool samples. These samples were then with a seed virus of EV71 strain FY7VP5/AH/CHN/2008 shipped to the Chinese National Institute for Food and (genotype C4, which is the predominant strain in Drug Control to test for IgM antibody, EV71-specifi c RNA, mainland China).5 Each dose of placebo contains 0·18 mg virus isolation,13 and VP1 amplifi cation. The data and safety of alum adjuvant and no EV71 antigen. monitoring committee classifi ed participants with EV71- Vaccine or placebo was administered intramuscularly associated disease to one of three groups (EV71-associated to the anterolateral side of the thigh (for participants HFMD, herpangina, or non-specifi c symptoms) according aged 6–11 months) or the deltoid muscle (those aged to signs and symptoms. 12–35 months) at day 0 and 28. Participants were Additionally, for each confi rmed case of EV71- observed for immediate adverse events for 30 min after associated disease we selected fi ve controls (matched each injection. Fever or any other adverse events were for location and age) without EV71-associated disease recorded by their guardians on diary cards within 28 days, and compared their neutralising antibody titre on day which were checked by investigators to assure complete- 56 with that of participants. We assumed that cases and ness and accuracy. We took blood samples from all participants at Panel 1: Primary endpoint defi nitions day 0 (immediately before the fi rst dose) and day 56 (28 days after second dose). In addition, we took blood Enterovirus 71 (EV71)-associated hand, foot, and mouth disease (HFMD) is characterised samples at month 8 and 14 from a subset of participants by febrile illness with papulovesicular rash (can occasionally be maculopapular without (immunogenicity cohort) by randomised cluster samp- vesicular lesion) on palms and soles, with or without vesicles or ulcers in the mouth, ling based on township for assessment of immuno- buttocks, knees or elbows,6 which is caused by EV71 virus (positive either for EV71 genicity and persistence (appendix p 5). Researchers at isolation or at least two consecutive EV71-specifi c RNA tests). the Chinese National Institute for Food and Drug Control EV71-associated disease is defi ned as clinical symptoms including HFMD, herpangina, measured neutralising antibody against EV71 with a neurological signs (aseptic meningitis or encephalitis) with or without serious sequelae, 5 modifi ed cytopathogenic eff ect assay. A data and safety and nonspecifi c illnesses—eg, febrile illness, viral exanthema, respiratory infection—which monitoring committee of nine members—including are caused by EV71 virus (positive either for EV71 isolation or at least two consecutive paediatricians, infectious disease doctors, epidemiol- EV71-specifi c RNA tests). ogists, and an immunologist—reviewed all the clinical, www.thelancet.com Vol 381 June 8, 2013 2025 Articles

their controls had similar exposure risk. All surveillance and a drop-out rate of 20% by the end of surveillance. A data were recorded in data management software sample size of 5000 participants per group would provide EV71VCTSM (version 1·30). The primary endpoints a 93·5% statistical power to show effi cacy with a lower were the occurrence of EV71-associated HFMD during bound of the 95% CI exceeding 20%. Vaccine effi cacy the surveillance period in the per-protocol population was defi ned as 100 × (1 – vaccine incidence density and EV71-associated disease during the surveillance rate / placebo incidence density rate). Incidence density period in the per-protocol population (panel 1). Secon- rate for each group was calculated as the number of cases dary end points for safety were occurrence of solicited divided by the total person-time of follow-up, censoring adverse events within 7 days of injection, any adverse for participants who had confi rmed EV71, died, or events within 28 days of injection, and all serious migrated away. Vaccine effi cacy was estimated by a Cox adverse events during the study period. Other secon- proportional hazard model. If no participant in the dary endpoints were geometric mean titre, serocon- vaccine group had EV71 disease, the 95% CI for effi cacy version rate, seropositivity rate, and geometric mean was assessed by an exact conditional procedure fold increase in each group. dependent on a Poisson distribution. We used the Kaplan-Meier method to estimate event-free survival. Statistical analysis We analysed a subset of participants for vaccine We calculated the required sample size with SAS immunogenicity and consistency between batches of (version 9·1) with the assumptions of 80% overall vaccine vaccine. 600 participants per group would provide at least effi cacy, EV71-associated HFMD incidence density rate of 95% power to show the superiority of vaccine over placebo eight cases per 1000 person-years in the placebo group, according to data from previous trials.5,12 200 participants

12 385 participants assessed for eligibility

2140 excluded* 582 did not meet inclusion criteria 1558 met some exclusion criteria

10 245 randomly assigned

205 did not receive second dose 5120 assigned to vaccine and received 5125 assigned to placebo and received 182 did not receive second dose 108 withdrew consent first dose first dose 104 withdrew consent 49 lost to follow-up 52 lost to follow-up 29 unsolicited AEs† 19 unsolicited AEs (2 SAEs)† 10 solicited AEs† 1 solicited AE† 8 excluded by investigators 5 excluded by investigators 1 refused to be vaccinated 4915 received second dose of vaccine 4943 received second dose of placebo 1 refused to be vaccinated

9 discontinued 10 discontinued 5 lost to follow-up 5 lost to follow up 3 migrated out 2 died (drowning) 1 onset before day 56 2 migrated out 1 onset before day 56

5111 included in surveillance 5115 included in surveillance beginning at day 56 beginning at day 56

294 discontinued follow-up 268 discontinued follow-up 206 lost to follow-up 179 lost to follow-up 68 withdrew consent 74 withdrew consent 18 migrated out 15 migrated out 2 died (1 intracranial haemorrhage, 1 acute 5111 included in efficacy analysis 5115 included in efficacy analysis lymphoblastic leukaemia) (modified intention to treat) (modified intention to treat) 4907 included in efficacy analysis 4939 included in efficacy analysis (per protocol)‡ (per protocol)§ 5117 included in safety analysis 5123 included in safety analysis 599 included in immunogenicity 620 included in immunogenicity subset subset 599 included in lot consistency analysis

Figure 1: Trial profi le The modifi ed intention-to-treat analysis included all participants who received at least one dose and entered surveillance at day 56. AE=adverse event. SAE=serious adverse event. *For details, see appendix p 6. †For details, see appendix p 39. ‡204 participants were excluded because they received only the fi rst dose. §175 participants received only the fi rst dose and one participant received the vaccine at second dose in the placebo group, so were excluded from the per-protocol analysis.

2026 www.thelancet.com Vol 381 June 8, 2013 Articles

per vaccine batch would provide at least 80% power to Vaccine group Placebo group detect consistency in vaccine-induced geometric mean Effi cacy analysis population (per protocol) titre with a pre-defi ned equivalence range (0·5–2·0) for n 5111 5115 geometric mean titre ratios. Mean age (months) We assessed correlates of immunity for all participants Overall 18·6 (8·1) 18·6 (8·1) with confi rmed EV71 and their controls who had data for In those aged 6–11 months 8·7 (1·7) 8·7 (1·7) antibody titre at day 56. We calculated s ensitivity In those aged 12–35 months 21·7 (6·6) 21·8 (6·7) (proportion of participants with EV71-associated disease Boys 2922 (57·2%) 2805 (54·8%) who have a titre less than the cutoff at day 56), specifi city Bodyweight (kg) 12·5 (2·2) 12·5 (2·2) (proportion of matched controls who have a titre equal or BMI 18·8 (2·5) 18·7 (2·4) greater than the cutoff at day 56), and corresponding Safety analysis population Youden index ([sensitivity + specifi city] – 1). The cutoff n 5117 5123 with the maximum Youden index could be thought of as Mean age (months) a surrogate of protection for it could provide the clearest 14 Overall 18·6 (8·1) 18·6 (8·1) distinction between cases and controls. In those aged 6–11 months 8·7 (1·7) 8·7 (1·7) To compare safety and immunogenicity endpoints we In those aged 12–35 months 21·7 (6·6) 21·8 (6·7) used Student’s t test for log-transformed antibody titres, 2 Boys 2927 (57·2%) 2813 (54·9%) and Fisher’s exact or χ for categorical data. Antibody titres Bodyweight (kg) 12·5 (2·2) 12·5 (2·2) lower than 1:8 were assigned values of 1:4 for calculation. BMI 18·8 (2·5) 18·7 (2·4) We did the primary analysis of vaccine effi cacy in the Immunogenicity analysis population per-protocol population, which consisted of all eligible n 599 620 participants who completed the two-dose regimen and Mean age (months) entered the surveillance period at day 56. We also assessed effi cacy in the modifi ed intention-to-treat Overall 18·6 (8·1) 18·5 (7·9) population, which consisted of all participants who In those aged 6–11 months 8·6 (1·7) 8·6 (1·7) received at least one dose and were followed up from In those aged 12–35 months 21·8 (6·6) 21·6 (6·4) day 56. We analysed safety in all participants who had Boys 343 (57·3%) 345 (55·6%) safety data available, and immunogenicity analyses were Bodyweight (kg) 12·3 (2·3) 12·3 (2·2) done in the per-protocol population of the immunogenic BMI 18·4 (2·6) 18·5 (2·8) subset of participants. Statistical analyses were done by Data are mean (SD), or n (%). BMI=body-mass index. external, independent statisticians with SAS (version 9.1). This study is registered with ClinicalTrials.gov, number Table 1: Baseline characteristics NCT01508247.

Vaccine group Placebo group Vaccine effi cacy Person-years Cases Incidence density Person-years Cases Incidence density % (95% CI) p value* at risk (n) rate (cases/1000 at risk (n) rate (cases/1000 person-years) person-years) Per-protocol population† EV71-associated disease 4725·4 8 1·7 4742·9 41 8·6 80·4% (58·2 to 90·8) <0·0001 EV71-associated HFMD 4725·4 3 0·6 4742·9 30 6·3 90·0% (67·1 to 96·9) 0·0001 EV71-associated other cases 4725·4 5 1·1 4742·9 11 2·3 54·3% (–31·4 to 84·1) 0·15 Modifi ed intention-to-treat population EV71-associated disease 4915·5 8 1·6 4903·5 44 9·0 81·9% (61·5 to 91·5) <0·0001 EV71-associated HFMD 4915·5 3 0·6 4903·5 33 6·7 90·9% (70·4 to 97·2) 0·0001 EV71-associated other cases 4915·5 5 1·0 4903·5 11 2·2 54·6% (–30·6 to 84·2) 0·14 Participants in per-protocol population with baseline titre <1:8 EV71-associated disease 3347·5 8 2·4 3345·4 40 12·0 80·0% (57·2 to 90·6) <0·0001 EV71-associated HFMD 3347·5 3 0·9 3345·4 29 8·7 89·6% (66·0 to 96·9) 0·0002 EV71-associated other cases 3347·5 5 1·5 3345·4 11 3·3 54·5% (–31·1 to 84·2) 0·14

EV71-associated other cases are EV71-associated disease with non-specifi c symptoms, including 14 cases of upper respiratory tract infection and two cases of diarrhoea caused by EV71 infection. EV71=enterovirus 71. HFMD=hand, foot, and mouth disease. *For comparison of incidence density rates between groups. †Three patients with EV71-associated disease were included in the modifi ed intention-to-treat population but excluded from the per-protocol population because they did not receive the second dose.

Table 2: EV71 vaccine effi cacy against EV71-associated diseases during 1 year of surveillance

www.thelancet.com Vol 381 June 8, 2013 2027 Articles

Role of the funding source immediately before the fi rst dose from 10 218 (99·7%; The sponsor helped to design the trial, but had no role in 5108 in the vaccine group vs 5110 in the placebo group) data collection, statistical analysis, interpretation, or participants and 28 days after second dose from writing of the report. All authors had full access to all 9896 (96·6%; 4949 vs 4947) participants. The proportion data. F-CZ, Z-LL, X-LS had primary responsibility and of participants who withdrew from the study between the fi nal decision to submit for publication. days 0 and 56 did not diff er signifi cantly between vaccine and placebo groups (appendix p 7). Results 9846 (96·3%) participants were eligible for the Figure 1 shows the trial profi le. 12 385 children were primary effi cacy analysis in the per-protocol population assessed for eligibility in January, 2012, of whom (4907 in the vaccine group vs 4939 in the placebo group). 10 245 (82·7%) were enrolled and randomly assigned to 10 226 participants (5111 in the vaccine group vs 5115 in vaccine (n=5120) or placebo (n=5125). 9858 (96·2%) the placebo group) were included in surveillance in participants received the second dose of vaccine (n=4915) the modifi ed intention-to-treat population. Dropout and placebo (n=4943). Blood samples were taken was low during the surveillance period: only 294 of 5111 (5·8%) participants in the vaccine group and A 268 of 5115 (5·2%) in the placebo group were lost to 100·0 follow-up. The safety analysis population consisted of 10 240 participants (5117 in the vaccine group vs 5123 in the placebo group). Five participants were excluded from the safety assessment because they had lost their diary cards. For the analysis of immunogenicity and antibody persistence we assessed 1219 participants at 99·5 day 56, 1134 at month 8, and 1123 at month 14. 599 participants who received EV71 vaccine and had data for antibody titre at day 56 were included in the analysis of batch consistency. Table 1 and appendix p 8 show baseline characteristics. Vaccine group Placebo group Surveillance lasted from March, 2012 to March, 2013,

Participants without EV71-associated disease (%) without EV71-associated Participants 99·0 during which time 7328 participants had 17 038 episodes

VE=81·9, 95% CI 61·5–91·5; p<0·0001 of illness (appendix p 9). 52 participants had laboratory- 0 confi rmed EV71-associated disease (appendix pp 10–15). 0 100 200 300 400 Time since day 56 (days) 51 of these participants were seronegative, with a base- Number at risk line neutralising antibody titre of less than 1:8, and one Vaccine 5111 5079 5034 4949 4809 Placebo 5115 5057 5021 4943 4803 participant had a titre of 1:8 at baseline. 46 (88·5%) of the 52 participants were positive in amplifi cation of VP1, B and phylogenetic analysis showed that all EV71 strains 100·0 were genotype C4 (appendix pp 16–17). Estimated vaccine effi cacy in the per-protocol population was 90·0% (67·1–96·9) for EV71-associated HFMD and 80·4% (58·2–90·8) for EV71-associated disease (table 2). Estimated vaccine effi cacy in the modifi ed intention-to- treat population was much the same as in the primary 99·5 analysis (table 2). Analysis of a modifi ed per-protocol population only including participants with baseline titre less than 1:8, showed much the same vaccine effi cacy as the normal per-protocol population. Figure 2 shows Kaplan-Meier cumulative event-free survival curves. No diff erence was noted between effi cacy in the

Participants without Participants EV71-associated disease (%) 99·0 fi rst 6 months and 1 year (appendix p 18). Appendix VE=80·4, 95% CI 58·2–90·8; p<0·0001 pp 19–20 shows analyses stratifi ed by age and centre. 0 0 100 200 300 400 Antibody and study centre were correlated with risk of Number at risk Time since day 56 (days) EV71-associated disease. Appendix pp 19–21 shows the Vaccine 4907 4879 4839 4760 4629 effi cacy adjusted for covariates. Eight patients in the Placebo 4939 4888 4854 4783 4655 placebo group (modifi ed intention-to-treat population) were admitted to hospital for EV71-associated HFMD Figure 2: Kaplan-Meier survival analysis of cumulative risk of not having EV71-associated disease Note, the y-axes are broken. For the modifi ed intention-to-treat population (A) and the per-protocol population (including one severe case of HFMD accompanied by (B). VE=vaccine effi cacy. EV71=enterovirus 71. encephalitis and oral herpes) compared with none in the

2028 www.thelancet.com Vol 381 June 8, 2013 Articles

vaccine group, suggesting a vaccine effi cacy of 100% Vaccine group Placebo group p value (95% CI 41·6–100·0) for potential severe (admitted to hospital) or severe cases (appendix p 24). All patients Day 0 recovered after admission to hospital. We noted no n 599 620 ·· herpangina alone caused by EV71. The EV71 season GMT (95% CI) 11·5 (9·9–13·3) 12·1 (10·4–14·1) 0·61 peaked between April and July, and trends in prevalence Proportion with titre ≥1:8 (n; %, 95% CI) 163 (27·2%, 23·7–31·0) 170 (27·4%, 23·9–31·1) 0·94 of EV71 diff ered between the four centres with the Proportion with titre ≥1:16 (n; %, 95% CI) 157 (26·2%, 22·7–29·9) 165 (26·6%, 23·2–30·3) 0·87 highest incidence in Baoying county (appendix p 25). Proportion with titre ≥1:32 (n; %, 95% CI) 147 (24·5%, 21·1–28·2) 161 (26·0%, 22·6–29·6) 0·57 Median periods of virus shedding were 13·8 days in the Day 56 vaccine group and 14·5 days in the placebo group n 599 620 ·· (p=0·63; appendix p 26). GMT (95% CI) 325·3 (284·8–371·7) 13·0 (11·1–15·2) <0·0001 Of the 17 038 episodes of illness recorded during GMFI (95% CI) 28·4 (25·5–31·6) 1·1 (1·0–1·1) <0·0001 surveillance, 1704 were clinically diagnosed as HFMD, Seroconversion (n; %, 95% CI) 549 (91·7%, 89·1–93·7) 19 (3·1%, 1·9–4·7) <0·0001 of which only 36 (2·1%) were fi nally confi rmed as being Proportion with titre ≥1:8 (n; %, 95% CI) 597 (99·7%, 98·8–100·0) 181 (29·2%, 25·6–33·0) <0·0001 associated with EV71, 577 (33·9%) were positive for Proportion with titre ≥1:16 (n; %, 95% CI) 596 (99·5%, 98·5–99·9) 172 (27·7%, 24·3–31·5) <0·0001 Coxsackie A virus 16, 588 (34·5%) were positive in other Proportion with titre ≥1:32 (n; %, 95% CI) 571 (95·3%, 93·3–96·9) 168 (27·1%, 23·6–30·8) <0·0001 enteroviruses, and 503 (29·5%) were not associated Month 8 with enterovirus. The incidences of coxsackie A virus 16, n 559 575 ·· other enteroviruses, and the overall number of clinical GMT (95% CI) 187·3 (162·4–216·0) 18·2 (15·1–21·8) <0·0001 cases of HFMD did not diff er signifi cantly between GMFI (95% CI) 16·3 (14·4–18·3) 1·5 (1·4–1·7) <0·0001 vaccine and placebo groups (appendix pp 27–30). Proportion with titre ≥1:8 (n; %, 95% CI) 556 (99·5%, 98·9–100·0) 223 (38·8%, 34·8–42·8) <0·0001 Baseline immunity—as measured by geometric mean Proportion with titre ≥1:16 (n; %, 95% CI) 542 (97·0%, 95·5–98·4) 183 (31·8%, 28·0–35·6) <0·0001 titre—was much the same in vaccine and placebo groups Proportion with titre ≥1:32 (n; %, 95% CI) 494 (88·4%, 85·7–91·0) 179 (31·1%, 27·4–34·9) <0·0001 (table 3, appendix p 31). Two doses of EV71 vaccine Month 14 elicited a substantial increase of neutralising antibody n 549 574 ·· titre compared with placebo. From day 56 to month 8, the GMT (95% CI) 191·9 (170·8–215·6) 16·2 (13·6–19·1) <0·0001 antibody titre in the vaccine group waned signifi cantly, GMFI (95% CI) 17·2 (15·1–19·7) 1·3 (1·2–1·4) <0·0001 and then remained consistent for the next 6 months. Proportion with titre ≥1:8 (n; %, 95% CI) 545 (99·3%, 98·6–100·0) 220 (38·3%, 34·4–42·3) <0·0001 Even by excluding potentially EV71-infected participants Proportion with titre ≥1:16 (n; %, 95% CI) 539 (98·2%, 97·1–99·3) 190 (33·1%, 29·3–37·0) <0·0001 who had a four-fold increase of antibody titre after day 56, Proportion with titre ≥1:32 (n; %, 95% CI) 508 (92·5%, 90·3–94·7) 181 (31·5%, 27·7–35·3) <0·0001 we noted no signifi cant drop in titre from month 8 to month 14 (appendix p 32). However, geometric mean Seroconversion is defi ned as pre-vaccination titre less than 1:8 and post-vaccination titre 1:32 or more, or titre in the vaccine group was signifi cantly higher pre-vaccination titre 1:8 or more and at least four-fold increase post-vaccination. GMT=geometric mean titre. GMFI=geometric mean fold increase. than that in the placebo group during the whole study (appendix pp 33–34). The ratios of geotmetric mean titre Table 3: Immune response to EV71 vaccine or placebo in the per-protocol population between any two vaccine batches at day 56 were in the predefi ned range (0·5–2·0), suggesting good consistency between production batches (appendix p 35). signifi cantly between groups (table 4). The most com- 137 of 10 240 (1·3%) participants had 139 serious mon symptoms at injection site were induration, adverse events between day 0 and month 14: 62 of erythema, and pain; the most common systemic symp- 5117 (1·2%) in the vaccine group and 75 of 5123 (1·5%) toms were fever and diarrhoea (table 4). Overall, much in the placebo group (table 4, appendix p 36). Although the same proportions of adverse events occurred within the incidence of serious adverse events between days 0 0–28 days after vaccination in the vaccine group and 56 was higher in the placebo group than in the (3644/5117; 71·2%) and placebo group (3603/5123; vaccine group (p=0·0112; table 4, appendix p 37), the 70·3%). 59 participants did not have the second dose diff erence was not signifi cant for the whole study because of adverse events (39 in the vaccine group vs 20 period. Four participants died (appendix p 38): two in in the placebo group; p= 0·0129). the vaccine group (from acute lymphoblastic leukaemia, To investigate correlates of immunity, we assessed intracranial haemorrhage) and two in the placebo group 51 participants with EV71-associated disease and (from drowning). None were deemed to be caused by 254 matched controls (one in each group was excluded vaccination. Within 7 days of injection, 2368 of 5117 because a blood sample at day 56 was unavailable). (46·3%, 95% CI 44·9–47·7) participants receiving Appendix p 40 shows the sensitivity, specifi city, and vaccine and 2250 of 5123 (43·9%, 42·6–45·3) receiving Youden index for each titre. The maximum Youden index placebo reported solicited adverse reactions (table 4). was provided by a titre of 1:16. Accounting for sensitivity, Although the incidence of solicited adverse reactions specifi city, and Youden index, a titre of between 1:16 was signifi cantly higher in the vaccine group (p=0·0165) and 1:32 seems to provide the best protection against during days 0–7, grade 3 events did not diff er EV71-associated diseases. www.thelancet.com Vol 381 June 8, 2013 2029 Articles

Vaccine group Placebo group p value Discussion (n=5117) (n=5123) Phase 1 and 2 trials have shown satisfactory safety and Serious adverse events immunogenicity of an investigational EV71 vaccine in Within 28 days after 10 (0·2%) 25 (0·5%) 0·0112 Chinese children (panel 2).5,12 Effi cacy of EV71 vaccine each dose* has not been reported worldwide. During active surveil- Within 0–14 months 62 (1·2%) 75 (1·5%) 0·27 lance, vaccine effi cacy was 90·0% against EV71-associated Deaths 2 (<0·1%) 2 (<0·1%) 1·00 HFMD and 80·4% against EV71-associated disease: Vaccine-related 0 (0·0%) 0 (0·0%) ·· similar effi cacies to other enterovirus-inactivated Solicited adverse reactions within 0–7 days vaccine—eg, inactivated poliovirus vaccine.15,16 Day 56 Any 2368 (46·3%) 2250 (43·9%) 0·0165 was planned to be the beginning of the surveillance Grade 3 98 (1·9%) 78 (1·5%) 0·13 period, when protection elicited by the EV71 vaccine Injection-site adverse reactions should have been at its peak. In fact, only two cases of Induration EV71-associated disease were reported between day 0 Any 293 (5·7%) 254 (5·0%) 0·08 and day 56, one in each group. Grade 3 1 (<0·1%) 0 (0·0%) 0·50 We did not measure protection from EV71 infection for Erythema several reasons. First, measurement of EV71 infection is Any 246 (4·8%) 247 (4·8%) 0·97 impractical because many people have asymptomatic Grade 3 2 (<0·1%) 3 (<0·1%) 1·00 infection, which means that all participants would need Pain to be periodically sampled and tested for virus or RNA.6 Any 165 (3·2%) 143 (2·8%) 0·20 Furthermore, seroconversion arising from vaccination Grade 3 0 (0·0%) 0 (0·0%) ·· cannot be distinguished from infection. Second, because Swelling EV71 infection does not always cause symptoms, Any 70 (1·4%) 65 (1·3%) 0·66 protection from infection is not necessarily the most Grade 3 0 (0·0%) 0 (0·0%) ·· important outcome. Furthermore, continuous circulation Pruritus of EV71 might provide natural boosting and help to Any 68 (1·3%) 57 (1·1%) 0·32 maintain immunity. Thus, at present, protection from Grade 3 2 (<0·1%) 0 (0·0%) 0·25 EV71-associated disease rather than infection should be Systemic adverse reactions the priority. Fever Although HFMD is the most typical clinical mani- Any 1799 (35·2%) 1738 (33·9%) 0·19 festation of symptomatic EV71 infection, several other 17,18 Grade 3 88 (1·7%) 71 (1·4%) 0·17 pathogens can also cause HFMD-like disease; further- Diarrhoea more, EV71 infection can cause various clinical mani- festations.19 Therefore, the causal relation between Any 309 (6·0%) 324 (6·3%) 0·55 EV71 and HFMD was complicated. We took several steps Grade 3 1 (<0·1%) 2 (<0·1%) 1·00 to improve specifi city and sensitivity. The Nation wide Decreased feeding Notifi able Infectious Diseases Reporting Infor mation Any 299 (5·8%) 311 (6·1%) 0·63 System in China is a passive surveillance system that is Grade 3 1 (<0·1%) 2 (<0·1%) 1·00 neither sensitive nor specifi c enough to detect all Crying EV71-associated cases. We used active surveillance of Any 290 (5·7%) 286 (5·6%) 0·85 many illnesses. We frequently contacted the guardians of Grade 3 2 (<0·1%) 1 (<0·1%) 0·62 participants and instructed them seek medical attention Nausea or vomiting if any symptoms were noticed. More than 1000 staff Any 185 (3·6%) 213 (4·2) 0·16 worked on the study site during the surveillance period, Grade 3 2 (<0·1%) 1 (<0·1%) 0·62 especially during the HFMD epidemic season. Throat Fatigue swabs, rectal swabs, and stool samples were collected for Any 97 (1·9%) 119 (2·3%) 0·13 more than 94·2% episodes of illness within 24 h of onset. Grade 3 0 (0·0%) 2 (<0·1%) 0·50 All specimens that were positive by real-time PCR assay Total adverse events 3644 (71·2%) 3603 (70·3%) 0·33 in local laboratories were retested and virus isolated within 28 days after each dose by the National Institute for Food and Drug Control Withdrawn due to 39 (0·8%) 20 (0·4%) 0·0129 in China. adverse events† Although EV71 was isolated from all participants with symptoms of EV71-associated disease, illness in Data are n (%) unless otherwise indicated. Grade 3 events were those deemed 16 participants who had non-specifi c symptoms—eg, severe—ie, prevented activity. *For further classifi cations see appendix p 37. †For further classifi cations see appendix p 39. upper respiratory tract infection and diarrhoea—might have been caused by coinfection with other pathogens Table 4: Serious adverse events and adverse reactions with incidence and not EV71. EV71 disease with non-specifi c signs— greater than 1% in the safety analysis population eg, fever, cough, diarrhoea—has been reported in

2030 www.thelancet.com Vol 381 June 8, 2013 Articles

epidemiological studies, but represents only a small proportion of all EV71 infections, usually around 4–8%.20,21 Panel 2: Research in context In our study, symptoms of upper respiratory tract Systematic review infection occurred in 14 of 52 (25%) participants, We searched PubMed with the terms “enterovirus 71”, “vaccine”, and “trial” with no date therefore the likelihood of co-infection—at least in some or language restrictions on April 27, 2013. Only fi ve immunogenicity and safety studies patients—is high. have been reported. Four were done in mainland China and one was done in Taiwan. No One limitation is that we did not test for pathogens previous phase 3 study of effi cacy of enterovirus 71 vaccine has been reported. other than EV71. If patients with non-specifi c disease were co-infected, specifi city might have decreased, Interpretation leading to underestimation of vaccine effi cacy.22 Because After a promising phase 2 trial, this study is the fi rst report of the clinical effi cacy of an HFMD is a characteristic symptom of enterovirus infec- inactivated enterovirus 71 vaccine. The vaccine provided signifi cant protection against tion, specifi city for detecting EV71 might have been enterovirus 71-associated disease, especially EV71-associated hand, foot, and mouth higher in the 36 patients with HFMD than in the disease in children aged 6–35 months. The experimental vaccine is at least as 15,16 16 patients with non-specifi c symptoms, which could immunogenic and effi cacious as traditional inactivated enterovirus vaccines. Our data explain the higher protective effi cacy against EV71- also provide the fi rst evidence of immunological correlates of protection against associated HFMD than against EV71-associated with enterovirus 71-associated disease. non-specifi c symptoms. All virus strains isolated from confi rmed cases in this study were genotype C4, the predominant strain in mainland China.1,6 However, in children. Although EV71 did not account for a high other regions, B4, B5, C2, and C5 have been reported.1,6,23 proportion of HFMD cases in this study, it has in other Variation between genotypes might compromise the settings—eg, in Sarawak28 and Taiwan.13 Large HFMD effi cacy of the vaccine. epidemics occur almost every year, whereas EV71 Most vaccines for communicable diseases have a direct epidemics occur at intervals of roughly 3 years in each eff ect on the vaccinated population and an indirect eff ect region as the number of susceptible children on the unvaccinated population—ie, widespread vaccin- accumulates, suggesting that diff erent predominant ation can provide herd immunity.24 We did not estimate pathogens might cause the yearly HFMD outbreaks. We the eff ect of herd immunity, but to reduce its eff ect as noted diff erences in prevalence of EV71 associated with much as possible, we enrolled a predefi ned proportion of diff erent regions and baseline antibody concentrations, children in every township (less than 40%). Thus, no providing an explanation for the disproportionately high more than 20% of target children or a lower proportion event rate in Baoying; therefore, effi cacy trials of EV71 of the general population would receive the vaccine, vaccine should be multicentre studies. lessening the indirect eff ect of herd immunity. This Infection with EV71 is of particular concern because it approach also helped to maintain the chance of re- can cause severe disease and even death in children;6,13 exposure to EV71 and as a result, maintain the durability the EV71 vaccine could help to prevent hospital admission of vaccine protection. A slight—although not signifi - and severe cases. However, few patients in our study cant—increase in geometric mean titre from month 8 to were admitted to hospital or had severe disease, therefore month 14 in the vaccine group suggests that some the estimates of effi cacy against EV71 in such patients participants were re-exposed to EV71. Large-scale had wide 95% CIs. population immunisation would reduce the circulation During active surveillance, incidence density of EV71 of EV71 and risk of re-exposure, resulting in a fall in diseases in children aged 6–11 months was not natural boosting and durability of immunity, as occurs substantially diff erent to that in children aged for measles and mumps vaccines.25,26 12–35 months (0·9 cases per 1000 person-years vs We did not do serological screening before enrolment 2·0 cases per 1000 person-years), suggesting that to exclude seropositive participants. Participants who children as young as 6 months should be vaccinated.13,29 were seropositive at baseline can be assumed to be The target age of vaccination (age 6–35 months) crosses immune and therefore an absence of infection during several routine vaccine times; therefore, further studies the trial would not be a result of vaccination. However, are needed to assess any eff ect of concomitant routine this eff ect was mitigated by taking a blood sample vaccination on EV71 vaccine effi cacy. from participants before the fi rst vaccination. EV71 is The safety profi le and immunogenicity of the not the only pathogen that caused HFMD. In the EV71 vaccine was clinically acceptable and consistent 1-year surveillance period, only a small proportion of with previous clinical trials (panel 2). Withdrawal because cases of HFMD were confi rmed as associated with EV71, of adverse events was higher in the vaccine group than in indicating that EV71 might not be the dominant pathogen the placebo group, mainly because of fever, suggesting in this HFMD season.27 Despite its high effi cacy for an increased risk of fever caused by vaccine. A sample preventing EV71-associated HFMD, the EV71 vaccine size of 5000 participants receiving EV71 vaccine is not might have little part in reducing the overall incidence of suffi cient to detect very rare adverse events after HFMD, even by universal mass immunisation of immunisation, so ongoing surveillance is needed. www.thelancet.com Vol 381 June 8, 2013 2031 Articles

Participants with EV71-associated disease had signifi - 9 Xu J, Qian Y, Wang S, et al. EV71: an emerging infectious disease cantly lower antibody titres at day 56 than matched vaccine target in the Far East? Vaccine 2010; 28: 3516–21. 10 Qiu J. Enterovirus 71 infection: a new threat to global public health? participants. Although a titre of 1:16 had the highest Lancet Neurol 2008; 7: 868–69. Youden index, a titre of 1:32 had very similar Youden 11 Meng FY, Li JX, Li XL, et al. Tolerability and immunogenicity of an index. Because a higher sensitivity provides better pro- inactivated enterovirus 71 vaccine in Chinese healthy adults and children: An open label, phase 1 clinical trial. tection from EV71-associated disease and the antibody Hum Vaccin Immunother 2012; 8: 668–74. titre fell over time, we suggest that a titre of 1:32 be used 12 Zhu FC, Wang JZ, Li XL, et al. Reactogenicity and Immunogenicity as a surrogate of protection against EV71-associated of an Enterovirus 71 Vaccine in Chinese Healthy Children and disease. This exploratory analysis on the correlates of Infants. Pediatric Infect Dis J 2012; 31: 1158–65. 13 Ho M, Chen ER, Hsu KH, et al. An epidemic of enterovirus 71 immunity is a substantial step forward; however, more infection in Taiwan. Taiwan Enterovirus Epidemic Working Group. evidence of its validity is needed. N Engl J Med 1999; 341: 929–35. 14 Barrett PN, Berezuk G, Fritsch S, et al. Effi cacy, safety, and Contributors immunogenicity of a Vero-cell-culture-derived trivalent infl uenza F-CZ, Z-LL, and X-LS designed the trial and contributed to critical review vaccine: a multicentre, double-blind, randomised, and revision of the report. F-CZ was principal investigator. Z-LL placebo-controlled trial. Lancet 2011; 377: 751–59. contributed to the study protocol and led the laboratory analyses. F-YM, 15 Robertson SE, Traverso HP, Drucker JA, et al. Clinical effi cacy of a J-XL, Y-TZ, HT, KC, and Y-MH collected data, supervised the study, and new, enhanced-potency, inactivated poliovirus vaccine. Lancet 1988; wrote and revised the report. H-MG, Z-YZ, Y-YD, Y-JC, L-YZ, Y-CL, 1: 897–99. N-MS, LL, S-GC, XA, Y-GJ, F-JL, YZ, L-WZ, and X-QC led and 16 WHO. Immunological Basis for Immunization - Module 6: participated in the site work, including recruitment, and data collection Poliomyelitis. Geneva, Switzerland: World Health Organization, and data interpretation. X-LL, Q-YM, FG, XW, XY, and Q-HC did 1993. http://whqlibdoc.who.int/hq/1993/WHO_EPI_GEN_93.16_ laboratory analyses, interpreted data, and searched the published work. mod6.pdf (assessed March 22, 2013). LP and HJ analysed data. All authors reviewed and approved the fi nal 17 Hu YF, Yang F, Du J, et al. Complete genome analysis of version of the report. coxsackievirus A2, A4, A5, and A10 strains isolated from hand, foot, and mouth disease patients in China revealing frequent Confl icts of interest recombination of human enterovirus A. J Clin Microbiol 2011; X-LL, Y-TZ, Q-HC, and X-LS are employees of the National Engineering 49: 2426–34. Research Center of Innovative Vaccine of Beijing Vigoo Biological. 18 Hu YF, Du J, Zhao R, Xue Y, Yang F, Jin Q. Complete genome The other authors declare that they have no confl icts of interest. sequence of a recombinant coxsackievirus B4 from a patient with Acknowledgments a fatal case of hand, foot, and mouth disease in Guangxi, China. We thank all the investigators from Jiangsu Provincial Center for J Virol 2012; 86: 10901–02. Disease Control and Prevention, National Institute for Food and Drug 19 Ooi MH, Wong SC, Lewthwaite P, Cardosa MJ, Solomon T. Clinical features, diagnosis, and management of enterovirus 71. Control, Baoying County Center for Disease Control and Prevention, Lancet Neurol 2010; 9: 1097–105. Chaoyang District Center for Disease Control and Prevention, Donghai 20 Kennett ML, Birch CJ, Lewis FA, Yung AP, Locarnini SA, Gust ID. County Center for Disease Control and Prevention, Pizhou County Enterovirus type 71 infection in Melbourne. Bull World Health Organ Center for Disease Control and Prevention, and Lianyungang City 1974; 51: 609–15. Center for Disease Control and Prevention who worked on the trial. 21 Chang LY, Tsao KC, Hsia SH, et al. Transmission and clinical We also thank Yuan-Zheng Qiu for his help in revising the report. features of enterovirus 71 infections in household contacts in References Taiwan. JAMA 2004; 291: 222–27. 1 Solomon T, Lewthwaite P, Perera D, Cardosa MJ, McMinn P, 22 Orenstein WA, Bernier RH, Hinman AR. Assessing vaccine effi cacy Ooi MH. Virology, epidemiology, pathogenesis, and control of in the fi eld. Further observations. Epidemiol Rev 1988; 10: 212–41. enterovirus 71. Lancet Infect Dis 2010; 10: 778–90. 23 Lee MS, Tseng FC, Wang JR, Chi CY, Chong P, Su IJ. Challenges to 2 Jee YM, Cheon DS, Kim K, et al. Genetic analysis of the VP1 region licensure of enterovirus 71 vaccines. PLoS Negl Trop Dis 2012; of human enterovirus 71 strains isolated in Korea during 2000. 6: e1737. Arch Virol 2003; 148: 1735–46. 24 Paul Y. Herd immunity and herd protection. Vaccine 2004; 22: 301–02. 3 Ho M. Enterovirus 71: the virus, its infections and outbreaks. 25 Poland GA, Jacobson RM. The re-emergence of measles in developed J Microbiol Immunol Infect 2000; 33: 205–16. countries: time to develop the next-generation measles vaccines? 4 Weng KF, Chen LL, Huang PN, Shih SR. Neural pathogenesis of Vaccine 2012; 30: 103–04. enterovirus 71 infection. Microbes Infect 2010; 12: 505–10. 26 Dominguez A, Torner N, Castilla J, et al. Mumps vaccine 5 Zhu FC, Liang ZL, Li XL, et al. Immunogenicity and safety of an eff ectiveness in highly immunized populations. Vaccine 2010; enterovirus 71 vaccine in healthy Chinese children and infants: 28: 3567–70. a randomised, double-blind, placebo-controlled phase 2 clinical 27 Legay F, Leveque N, Gacouin A, et al. Fatal coxsackievirus A-16 trial. Lancet 2013. published online March 23. http://dx.doi. pneumonitis in adult. Emerg Infect Dis 2007; 13: 1084–86. org/10.1016/S0140-6736(12)61764-4. 28 Chan LG, Parashar UD, Lye MS, et al. Deaths of children during 6 WHO. A guide to clinical management and public health response an outbreak of hand, foot, and mouth disease in Sarawak, for hand, foot and mouth disease (HFMD). Geneva, Switzerland: Malaysia: clinical and pathological characteristics of the disease. World Health Organization, 2011. http://www.wpro.who.int/ For the Outbreak Study Group. Clin Infect Dis 2000; 31: 678–83. publications/docs/GuidancefortheclinicalmanagementofHFMD.pdf 29 Chang LY, King CC, Hsu KH, et al. Risk factors of enterovirus (accessed Sept 28, 2012). 71 infection and associated hand, foot, and mouth disease/ 7 Lum LCS, Wong KT, Lam SK, Chua KB, Goh AYT. Neurogenic herpangina in children during an epidemic in Taiwan. Pediatrics pulmonary oedema and enterovirus 71 encephalomyelitis. Lancet 2002; 109: e88. 1998; 352: 1391. 8 Chang L-Y, Lin T-Y, Hsu K-H, et al. Clinical features and risk factors of pulmonary oedema after enterovirus-71-related hand, foot, and mouth disease. Lancet 1999; 354: 1682–86.

2032 www.thelancet.com Vol 381 June 8, 2013